Online fault detection methods for chillers combining extended kalman filter and recursive one-class SVM

A novel hybrid method is proposed to detect faults for chiller subsystems without any faulty training data available, i.e. by training the normal data only and requires less feature variables compared to existing works.